Shadowing Properties of Optimization Algorithms

Abstract

Ordinary differential equation (ODE) models of gradient-based optimization methods can provide insights into the dynamics of learning and inspire the design of new algorithms. Unfortunately, this thought-provoking perspective is weakened by the fact that, in the worst case, the error between the algorithm steps and its ODE approximation grows exponentially with the number of iterations. In an attempt to encourage the use of continuous-time methods in optimization, we show that, if some additional regularity on the objective is assumed, the ODE representations of Gradient Descent and Heavy-ball do not suffer from the aforementioned problem, once we allow for a small perturbation on the algorithm initial condition. In the dynamical systems literature, this phenomenon is called shadowing. Our analysis relies on the concept of hyperbolicity, as well as on tools from numerical analysis.

Cited by in corpus (8)

• Continuous-in-Depth Neural Networks
• Continuous vs. Discrete Optimization of Deep Neural Networks
• A Continuous-time Perspective for Modeling Acceleration in Riemannian Optimization
• The connections between Lyapunov functions for some optimization algorithms and differential equations
• On The Convergence of Euler Discretization of Finite-Time Convergent Gradient Flows
• A Dynamical Systems Approach for Convergence of the Bayesian EM Algorithm
• Revisiting the Role of Euler Numerical Integration on Acceleration and Stability in Convex Optimization
• Continuous-time Models for Stochastic Optimization Algorithms